Skip to content
Demand forecasting for shops that don't have a data team
Decision Science

Demand forecasting for shops that don't have a data team

All articles
Decision Science August 3, 20262 min read

Demand forecasting for shops that don't have a data team

You do not need a data scientist to forecast demand. You need last year, the weather, the calendar, and the honesty to write the number down before the week happens — then check it after. Here is the practical version, and where the software earns its keep.

Jaswant Singh

Jaswant Singh

Co-Founder, CTO & COO, Kauzio

Every shop already forecasts. The owner who orders "the usual, plus a bit because it's getting warmer" is running a forecasting model — it just lives in their head, never gets written down, and never gets checked against what actually happened. The difference between that and what a supermarket chain does is not intelligence. It is discipline plus memory.

The forecast you can run on paper

A workable demand forecast for a small shop needs four inputs, all of which you already have.

The same period last year. Not last week — last year. Retail and hospitality demand rhymes annually far more than it trends weekly. Last August is a better guide to this August than last month is.

The recent trend. Is the last month running above or below the same month last year, and by roughly how much? Apply that ratio to what last year says next week should look like.

The calendar. Bank holidays, school terms, local events, paydays. These move demand far more than most owners' instincts admit, and they are all knowable in advance.

The weather. Not perfectly — but a heatwave forecast on Thursday should change what a café orders on Tuesday, and it is free information.

Write the number down. That is the step almost everyone skips, and it is the one that matters, because a forecast you never wrote down can never be wrong — which means it can never get better.

Where software actually earns its place

The paper version breaks in two places. It cannot watch every product at once, and it cannot remember how wrong it was in a way that improves the next guess. That is the honest pitch for tools like ours: Kauzio runs that same logic — seasonality, trend, calendar, weather — across every product you stock, continuously, and then does the part humans hate: it records what it predicted, waits, compares the prediction with what your till actually says, and adjusts.

We hold ourselves to the writing-it-down rule too. Every forecast Kauzio makes is sealed before the outcome is known, so the accuracy number you see is the real one, failures included. A forecast you can quietly forget is a forecast you cannot trust.

Start smaller than feels sensible

Do not try to forecast everything. Pick your ten most valuable products — the ones where being wrong costs real money in either direction, wasted stock or empty shelves — and forecast only those for a month. Ten written-down numbers, checked weekly, will change how you order more than any dashboard.

The goal is not a perfect forecast. Nobody has one. The goal is a forecast that is written down, checked, and slightly less wrong every month. That compounds.

#forecasting#demand#decision intelligence

Read next